Abstract
Attempts to reintroduce threatened species from ex situ populations (zoos or predator-free sanctuaries) regularly fail because of predation. When removed from their natural predators, animals may lose their ability to recognize predators and thus fail to adopt appropriate antipredator behaviors. Recently, northern quolls (Dasyurus hallucatus; Dasyuromorpha: Dasyuridae) conserved on a predator-free “island ark” for 13 generations were found to have no recognition of dingoes, a natural predator with which they had coevolved on mainland Australia for about 8,000 years. A subsequent reintroduction attempt using quolls acquired from this island ark failed due to predation by dingoes. In this study, we tested whether instrumental conditioning could be used to improve predator recognition in captive quolls sourced from a predator-free “island ark.” We used a previously successful scent-recognition assay (a giving-up density experiment) to compare predator-scent recognition of captive-born island animals before and after antipredator training. Our training was delivered by pairing live predators (dingo and domestic dog) with an electrified cage floor in repeat trials such that, when the predators were present, foraging animals would receive a shock. Our training methodology did not result in any discernible change in the ability of quolls to recognize and avoid dingo scent after training. We conclude either that our particular training method was ineffective (though ethically permissible); or that because these quolls appear unable to recognize natural predators, predator recognition may be extremely difficult to impart in a captive setting given ethical constraints. Our results point to the difficulty of reinstating lost behaviors, and to the value of maintaining antipredator behaviors in conservation populations before they are lost.
Globally, conservation managers are struggling to combat the growing species extinction crisis. To mitigate declines and extinctions, species increasingly are being placed into ex situ management in predator-free sanctuaries and zoos (Zippel et al. 2011; Pritchard et al. 2012). While ex situ conservation often is the only management action available to halt extinction, it also results in the loss of traits important to survival in situ (Jamieson and Ludwig 2012; Moseby et al. 2015; Muralidhar et al. 2019). Reintroductions from predator-free havens (e.g., from captivity, predator-free exclosures, or islands) are seen as increasingly important for species management, but a large proportion of such reintroductions fail (Griffith et al. 1989; Fischer and Lindenmayer 2000; Seddon et al. 2007). This failure is very frequently due to predation by both native and invasive predators (Fischer and Lindenmayer 2000; Jule et al. 2008; Moseby et al. 2015). Unless we can maintain or restore traits—such as predator aversion—that are critical to in situ fitness, reintroductions will repeatedly face a losing battle with predators. This growing realization has led to intense interest in whether antipredator training can be used to teach individual predator-naïve animals to recognize and avoid predators (Griffin et al. 2000; Moseby et al. 2012; Teixeira and Young 2014; Paulino et al. 2018; Ross et al. 2019).
In Australia, the invasion of toxic cane toads (Rhinella marina) has caused dramatic declines in a number of large, vulnerable native predators (Shine 2010). Like other Australian species with no coevolutionary history with bufonid toads, northern quolls lack physiological resistance to toad toxins (Ujvari et al. 2013) and rapidly die if they attempt to eat these toxic invaders. This vulnerability has resulted in dramatic declines and local extinctions of northern quolls from many parts of their former range (Burnett 1997; Woinarski et al. 2011). In 2003, insurance populations of quolls were established on two toad- and predator-free islands in Arnhem Land, Northern Territory, with the long-term aim of using these populations to reestablish mainland populations decimated by toads (Rankmore et al. 2008). These island populations were founded by quolls sourced from locations across the Top End of the Northern Territory, including Kakadu National Park (Rankmore et al. 2008). Quolls thrived on these islands, with an estimated population of 2,193 female northern quolls on Astell Island by 2014 (Griffiths et al. 2017). In 2016, northern quolls were collected from Astell Island, trained to avoid toads via conditioned taste aversion (O’Donnell et al. 2010), provided with rudimentary antipredator training, and reintroduced to Kakadu National Park (Jolly et al. 2018a). However, they did not survive long-term because of predation by dingoes (Canis familiaris dingo—Jolly et al. 2018a); a predator with which quolls coevolved for about 8,000 years (Cairns and Wilton 2016; Zhang et al. 2020). Despite having been isolated for only 13 generations on Astell Island, quolls from this predator-free “island ark” failed to recognize the scent of dingoes (Jolly et al. 2018b). A failure to recognize predators is the highest level of predator naïveté possible in a cascade of antipredator responses (Banks and Dickman 2007; Carthey and Banks 2014): without recognition, there can be no appropriate response.
The success of future reintroductions of northern quolls will depend upon improving their predator recognition and response. More generally, while we are increasingly becoming aware that antipredator behaviors are being lost in havens, we are less certain whether they can be reacquired. Training has been shown to improve the antipredator behaviors of some predator-naïve prey (Magurran 1989; McLean et al. 1999, 2000; Griffin et al. 2000, 2002), particularly when live predators are used to deliver the training (van Heezik et al. 1999; West et al. 2018; Blumstein et al. 2019; Ross et al. 2019). Such antipredator training may provide a general means of improving reintroduction success from ex situ population sources.
In 2016, we attempted to train wild-caught “island ark” northern quolls to avoid both cats and dingoes by employing a predator training method that paired a picture of each predator and each predator’s fur with a shock (snap trap; see Jolly et al. 2018a for details). Although the efficacy of this training method to elicit a behavioral change in quolls was not formally tested, dingoes were primarily responsible for the failure of this reintroduction, so this training method was clearly unable to meaningfully improve post-release survival. In the present study, we tested whether a different method of instrumental conditioning could be used to improve predator recognition in captive northern quolls sourced from the predator-free “island ark” population on Astell Island. We used a previously successful scent-recognition assay (a giving-up density [GUD] experiment—Jolly et al. 2018b) to compare predator-scent recognition of captive-born, Astell Island animals before and after antipredator training. Our training was delivered by pairing live predators (dingo and domestic dog) with an electrified cage floor, such that animals spending time on the ground would receive a shock when the predator was present. We predicted that if quolls learn to associate aversive cues with dingoes, this should be reflected in their recognition of dingo scent after antipredator training.
Materials and Methods
Study species
Northern quolls are a rabbit-sized, native Australian marsupial predator that occurred historically across much of northern Australia—from southeast Queensland to the tip of Cape York Peninsula, throughout the Top End of the Northern Territory, and the Kimberley and Pilbara regions of Western Australia (Van Dyck et al. 2013). They are generalist carnivores that consume both invertebrate and vertebrate prey (Oakwood 1997; Pollock 1999; Dunlop et al. 2017). Northern quolls are the largest semelparous marsupials (Oakwood et al. 2001). Both sexes mature at 11 months and male die-off typically occurs soon after reproduction (Dickman and Braithwaite 1992; Oakwood et al. 2001). After birth, juvenile northern quolls typically spend the first 4–5 months with their mother (Oakwood 2000).
Training population
We measured the antipredator responses of captive-born northern quolls whose parents were wild-caught from Astell Island—an insurance population established to mitigate against the threat of quoll extinction from the mainland due to the arrival of toxic invasive cane toads into the Northern Territory. This island population of quolls was established in 2003 by translocating 45 individuals from the Northern Territory mainland to Astell Island (−11.884, 136.418), North East Arnhem Land, Northern Territory. The founders of this population came from five localities from the northwestern Top End region, Northern Territory (Rankmore et al. 2008; Cardoso et al. 2009). Extensive monitoring of Astell Island quolls showed that the population grew very rapidly before overshooting its carrying capacity and declining to a stable size prior to collection for reintroduction (Griffiths et al. 2017). In 2016, we captured quolls from Astell Island for breeding at the Territory Wildlife Park, Berry Springs, Northern Territory. Some quolls were retained for breeding, while others (n = 29) were used in a wild-to-wild reintroduction to Kakadu National Park (Jolly et al. 2018a). By early 2017, there were cohorts of wild-caught, adult quolls, as well as their captive-born, adult offspring in captivity at the wildlife park. All the quolls used for predator training in this study were captive-born quolls from Astell Island parents and all were recently weaned and independent of their mothers.
Predator-free Astell Island
Astell Island was determined to be entirely quoll predator-free based on record data collected from extensive flora and fauna surveys of the island prior to the release of quolls in 2003 (Woinarski et al. 1999). There are no mammalian predators on the island, nor are there any avian or reptilian predators that would constitute a significant or sustained predation threat to northern quolls (Jolly et al. 2018b). Quolls collected from Astell Island were known to neither recognize nor avoid the scent of predators (dingoes and cats) that their mainland conspecifics both recognized and avoided (Jolly et al. 2018b). This was true both for wild-caught and captive-born Astell Island quolls (Jolly et al. 2018b).
Husbandry of northern quolls
All northern quolls were maintained individually in aviary-style enclosures in a quarantine facility at the Territory Wildlife Park. These enclosures were designed and built specifically for the park’s northern quoll breeding program. Quoll enclosures were of varying sizes but typically were 2 m × 4 m × 2.5 m (width, depth, and height). By alternating the placement of males and females, no quolls of the same sex were housed in adjoining enclosures. All enclosures were situated within a quarantine area isolated from the rest of the wildlife park, which is bordered by a 3-m predator- and toad-proof boundary fence, the perimeter of which is offset from quoll enclosures by at least 20–50 m. Quolls were moved to experimental enclosures for antipredator training. The quolls used in this study were born in this breeding facility and had never previously encountered canids. Quolls were fed a base diet of dry dog food, supplemented with whole mice, small whole fish, crickets, mealworms, and various fruits. Our study followed ASM guidelines (Sikes et al. 2016) and all applicable international, national, and/or institutional guidelines for the use of animals were followed. The University of Technology, Sydney Animal Ethics Committee approved the experimental design (ID number: 2015000175).
GUD experiment
The predator-scent recognition of all individual quolls was assayed using a GUD experiment repeated over three consecutive nights both before and after predator training at the Territory Wildlife Park between January and April 2017. This assay revealed large differences in predator-scent recognition between island and mainland quolls (Jolly et al. 2018b), we therefore used that response as a baseline for our expectations in this study. Each GUD box presented to quolls in each trial contained 2 liters of sterile wood shavings and 10 mealworms. A circular opening (15 cm diameter) in the top of each GUD box was fitted with a predator fur-filled fly-mesh collar (Fig. 1). Collars were filled either with dingo fur, cat fur, or no fur (control; Fig. 1). Fur was used as an olfactory cue instead of predator secretions because it provides a more reliable cue of immediate predator presence (Apfelbach et al. 2005). Dingo fur was sourced from an adult male and adult female dingo. Both dingoes were captive animals that were rescued from the wild as puppies. Fur was collected by shaving it directly from the dingoes immediately prior to each experimental trial. Fur from both dingoes was combined and used in each experimental trial. Cat (Felis catus) fur was collected from a number of adult male and female, domestic and feral cats (n ≈ 15). Cat fur was collected by veterinarians from The Ark Animal Hospital during routine procedures that required fur to be shaved from cats. Fresh cat fur was collected, and all sources combined, prior to the commencement of each experimental trial. To avoid cross contamination of scents, collars were prepared in separate rooms prior to each trial, and each collar was only used once.
Giving-up density boxes. Each northern quoll (Dasyurus hallucatus) was provided with three boxes, each of which had a predator-scented (dingo-fur and cat-fur) or control collar lining the inside of their access holes. Inside each box were 2 liters of sterile wood shavings and 10 mealworms.
During behavioral trials, quolls were presented with one of each of the three predator-scented GUD boxes and were allowed to investigate and feed from the boxes throughout the night of the trial. The next morning, the remaining mealworms were counted. Before predator training, each quoll was presented with GUD boxes on three consecutive nights, but the arrangement of the boxes was changed each night (e.g., Night 1: control-dingo-cat; Night 2: dingo-cat-control; Night 3: cat-control-dingo). Between trials all wood shaving and mealworms were replaced. The pre-training behavioral trials were completed for each quoll two nights prior to antipredator training, and post-training trials commenced the night after the final antipredator training session to determine whether there was an observable change in quoll behavior that may have been a result of learned predator aversion.
Antipredator training
Using an instrumental conditioning procedure, we attempted to train adult female captive-born, predator-naive Astell Island quolls (n = 17) to recognize and avoid dingo-scented GUD boxes. All experiments were conducted in an experimental bank of nine aviary-style enclosures (Fig. 2) that were isolated from the quoll’s home enclosures, but still located within the quoll quarantine area of the wildlife park. Quolls were moved, along with their individual hide box, from their home enclosure to smaller experimental enclosures (0.8 m wide × 2.5 m deep × 3 m tall) two nights prior to the trials. Each quoll was housed individually during training experiments. Experimental enclosures were constructed of 1 cm × 1 cm aviary mesh, with a corrugated iron roof. The perimeter wall of the aviary bank was lined with shade cloth, but there was no sensory obstruction between the individual quoll enclosures and the enclosure vestibule. Since there only were nine experimental enclosures, nine quolls were trained during the first training session; eight were trained in a second training session. Between sessions, experimental enclosures were thoroughly washed and disinfected.
Experimental setup for antipredator training. Live stimulus animals (dingo and cattledog) were presented in the vestibule of the enclosure bank behind a wire mesh barrier. Wires only were electrified during the 2-h period the dingo was presented in the vestibule. Responses of quolls were filmed by GoPro cameras during the 2-h training period. All quoll enclosures were occupied by quolls during the training period.
To elicit an avoidance response to dingo scent, we paired the introduction of a live dingo and cattledog (companions) to the vestibule of the experimental enclosure bank with the floor of the cage modified to give a mild electric shock (Speedrite AN20 energizer; maximum output 0.04 J; Fig. 2). The wires were placed such that quolls would be required to make contact with the wires to reach their food, which was provided on the ground at the front of each enclosure during the entire experiment. Following the pre-training GUD experiment, we instituted a week-long “training period” (seven nights), where, on four random nights, the dingo and cattledog were placed in the vestibule an hour after dark. The wires were electrified only during the 2 h that the canids were present in the vestibule of the quoll enclosures. The arrival of the canids coincided with the quolls’ usual feeding time (quolls were provided with their regular diet immediately prior to the presentation of the canids). To increase the proximity and likelihood of interactions between the canids and quolls, and quolls and electric wires, neither of the canids were fed on the day of the training, both were used to being fed the same dog biscuits presented to quolls but out of reach from the canids (i.e., the canids were actively trying to access food that was inside the quoll enclosures) and quolls were fed slightly less than normal the night before the training events. Any quolls that emerged from hiding while the canids were present in the vestibule would have been able to see, hear, and smell, these potential predators throughout the “training period.” In addition, the dingo had been presented previously with northern quoll carcasses on a number of occasions to stimulate a predatory association with quolls as prey. The dingo was allowed to play with and eat these carcasses. Although theoretically a pet, this dingo was not domesticated and was very aggressive toward smaller mammals, including smaller dogs, and would regularly and rapidly attack anything it perceived as alive and potential prey. For this reason, and because of the experimental enclosure design, the emergence of a quoll from its hide to forage at ground level would very likely have triggered an investigatory, if not predatory, response from this dingo. When the canids were not present, the electric wires were turned off, but the wires remained in their enclosures and quolls needed to make contact with these nonelectrified wires to feed. Canids were paired with the electrification of the wires on four, nonconsecutive, random nights during the training period. On the three nights during the training period that the canids were not present, everything remained the same (e.g., food was present to the quolls placed between the wires), but the canids were absent, and the wires were not electrified. GoPro cameras were used to record quoll responses to the canids. To be included in the study, quolls were required to be observed to approach their food, attempt to feed, and get shocked on at least one of the three nights when canids were present. After being shocked, most quolls appeared to rapidly retreat to their elevated hide box after interacting with the training apparatus.
Statistical analysis
For each individual quoll, we calculated the difference in the number of mealworms consumed after training minus the number consumed before training for each of the three predator-scent types (control, dingo, and cat). For each quoll, we then calculated the mean after versus before response across the three trials. Once our data were confirmed to be normally distributed and the variance homogenous, we used a one-way ANOVA to compare the after versus before response of quolls among the three scent types. This analysis was carried out using R (R Core Team 2019).
We benchmarked our results against data from predator-savvy, wild-caught, female quolls (n = 8) from the mainland (Jolly et al. 2018b). The behavioral responses of predator-savvy quolls were measured using the same assay, with exactly the same experimental design, but these quolls had been caught from areas where quolls likely have persisted alongside dingoes for about 8,000 years (Cairns and Wilton 2016; Zhang et al. 2020). We reasoned that the pattern of predator-scent recognition displayed by these mainland quolls allows them to detect and avoid predation events and, ideally, a similar pattern of behavior would be displayed by quolls that were successfully trained to avoid dingoes.
Results
Despite all quolls having been observed to be shocked by the electric stimuli while attempting to forage in the immediate presence of the dingo (at least once during the training period), there was no observable before/after difference in the mean numbers of mealworms consumed from the GUD boxes regardless of scent type (one-way ANOVA; F2,48 = 0.157, P = 0.855; Fig. 3). There did, however, appear to be some reduction in the willingness of quolls to forage in the presence of canids across the training period. On the first night of training 15 of 17 quolls (0.88), on the second night 12 of 17 quolls (0.71), and on the third night 10 of 17 quolls (0.59) emerged and were shocked by the electric stimuli. Most quolls were shocked on two of three training nights (0.82), four quolls were shocked on three of three training nights (0.24), and only a single quoll was only shocked once (0.06). This apparent reduction in willingness to forage in the presence of canids did not cross over into a willingness to avoid dingo scent in GUD trials.
Mean difference between the number of mealworms eaten by individual northern quolls from giving-up density (GUD) boxes of each scent type before and after predator training. Boxes show median and interquartile range (IQR) of the raw data; whiskers represent 1.5 times the IQR, or the range of the data, whichever is smaller. A negative number of mealworms eaten by quolls from dingo-scented boxes versus control-scented boxes after training would signify a shift in behavior suggesting they had acquired some learnt aversion to dingo scent.
We benchmarked our results against the pattern of response observed in wild-caught, predator-savvy quolls (Fig. 4). Against this reference, our post-training animals ate a similar proportion of mealworms from control boxes, but a substantially higher proportion of mealworms in the predator-scented boxes.
Mean proportion (± SE) of mealworms eaten by wild-caught, mainland female quolls (n = 8) and captive-bred, “island ark” female quolls (n = 17) from giving-up density boxes fitted with one of three treatment scent types during a 2-h period.
Discussion
Our results suggest that evolutionarily lost antipredator recognition may not be easily reinstated with the method of antipredator training we employed, despite us using the well-worn route of instrumental conditioning to achieve learning in animals (Thorndike 1911; Skinner 1948; Griffin et al. 2000). After undergoing training in which live predators (a dingo and domestic dog) were paired with an electric shock, there was no before/after training difference in the mean number of mealworms eaten by captive-bred, “island ark” northern quolls from each of the three types of scented GUD boxes (control, dingo-scented, and cat-scented; Fig. 3). That is, there was no observable change in the predator-scent recognition behavior of captive-born Astell Island quolls elicited by our antipredator training method. A previous release of quolls from Astell Island to Kakadu National Park ultimately failed because of predation by dingoes (Jolly et al. 2018a). The ultimate aim of our study was to investigate whether antipredator training improves post-release survival. Because there was no observable change in predator-scent aversion in predator-trained quolls, we had no reason to believe that the outcome of an additional reintroduction was likely to be different. We therefore ceased further reintroduction attempts for ethical reasons.
The results of our study show fairly conclusively that the cohort of predator-naïve quolls we attempted to train did not learn to associate an aversive event (electric shock) with the scent of dingoes. Our training method may have failed because it did not create a tight temporal association between predator cues and aversive stimulus, and such tight association is the most powerful way to effect instrumental learning (Griffin et al. 2000, 2001). Without actually allowing the canids into the cage with the quolls (a manipulation not available to us on ethical grounds), tight temporal pairing was effectively impossible in our case. To mitigate that, we ran multiple trials with and without cue and stimulus. Thus, our seven nights’ training may not have been sufficient time for quolls to correlate predator with aversive stimulus. It also is possible that electric shocks were less likely to be associated with a natural experience with a predator, and so may have been less useful for training (Griffin et al. 2000). We also cannot rule out whether potentially more salient cues, such as the sight and sound of the canids, were the cues that quolls learned to associate with the aversive stimulus. As the canids foraged around the vestibule during the training period, they inevitably made noises and movements that quolls would have been able to perceive. If auditory and/or visual cues were more salient than the olfactory cues given off by the canids, there is a possibility that these may have overshadowed the olfactory cues that we later used when testing whether quolls learned predator aversion. In addition, it is possible that the scent of the canids remained after they were removed from the vestibule following the end of the training period, potentially weakening the association between predator scent and perceived danger. When multiple cues are presented together, some or all can lose associative strength and may become less salient than if they were presented alongside the negative experience independently (Pearce 2013). Clearly it is not possible to isolate the cues given off by a live predator if used for antipredator training; however, future studies attempting to train predator aversion in predator-naïve animals would benefit by assessing all cues against which learning may have occurred. Such studies could benefit by pairing appropriate stimuli with conspecific alarm calls (or odors), or using trained demonstrators. In many animals, socially acquired fear responses toward predators can be rapidly learned when visual cues are paired with appropriate alarm calls (Griffin 2004). Young quolls spend extended periods foraging with their mothers, so it is possible that social learning is involved with acquisition of antipredator responses in this species.
Nonetheless, if we use recognition of predator scent as a measurable baseline, it is clear that our training attempts in this case failed to improve scent recognition toward the levels we observe in predator-savvy populations. Predator-savvy quolls clearly recognize and avoid the subtle cues that identify a predator’s potential presence (i.e., scent), rather than simply responding to the actual presence of a predator (and so a potentially immediate predation event). By its very presence, this behavioral aversion suggests that scent recognition is important for avoiding encounters with predators well in advance of an actual predatory encounter. That our training method failed to impart this scent-recognition speaks against it as a training method even if it has imparted recognition on some more proximal but unmeasured axis of sense (i.e., sight, sound).
It is entirely possible that our training method was simply inadequate; but there is an alternative hypothesis worth flagging: it may not be possible to train antipredator behaviors into animals that have suffered from the evolutionary loss of predator recognition (Griffin et al. 2000; Jolly et al. 2018b). The training methodology we employed (i.e., instrumental conditioning) has a long history of being used successfully to elicit learnt behaviors in animals (Thorndike 1911; Skinner 1948; Griffin et al. 2001, 2002), so we might have expected it to work here also. There also is mounting evidence that some animals can and do learn from antipredator training via training methods similar to those used here (van Heezik et al. 1999; Griffin et al. 2001; Moseby et al. 2012, 2015). Yet, our quolls showed no change in behavior when presented with dingo fur after training with a live dingo paired with an electric shock (Fig. 3), and their responses were starkly different to those of quolls that recognized dingo scent and avoided it (Fig. 4). Given that instrumental conditioning has support in its ability to promote learned aversion in animals both from the behavioral ecology and psychological literature (Griffin et al. 2000), it is worth considering that our quolls were inherently difficult to train because they lacked the fundamental capacity to recognize predators. Predator naïveté is thought to exist on a three-level scale of severity, with prey that demonstrate no ability to recognize a predator (and so adopt no antipredator behavior because of this) being the most severely affected (Banks and Dickman 2007). We suspect that animals suffering from this most severe form of predator naïveté (as do our island quolls) may be challenging to train. If an animal simply does not recognize when a predator is present or absent, the presence and absence of the aversive stimulus is not correlated with anything the animal observes.
If antipredator traits are not easily returned to predator-free populations by means of antipredator training, we may require an evolutionary means of returning them (i.e., natural selection—Moseby et al. 2016, 2018). That is, the selection pressures leading to loss of antipredator traits in predator-free havens would need to be reversed, and the simplest means of achieving this would be to return sustainable levels of predation to the population. Such a strategy would also generate real, uncontrolled predator encounters from which individuals might learn (or be preyed upon). Recently, conservation managers have been having success instilling antipredator behaviors into populations that lack or have lost these behaviors by pairing predators and prey in semi-wild, fenced sanctuaries (Blumstein et al. 2019). When paired with low densities of feral cats, both burrowing bettongs (West et al. 2018) and greater bilbies (Ross et al. 2019) have been shown to have improved antipredator behavior compared to control, predator-naïve individuals and, in bilbies, experience improved survival after introduction to areas with established cat populations (Blumstein et al. 2019; Ross et al. 2019). These results likely stem from a combination of learning and natural selection generated by the presence of predators.
While this large-scale predator-exposure approach looks encouraging, it inevitably results in injuries and death of animals (through predation) and so raises ethical concerns. We must, of course, consider the ethics and the logistics of large-scale predator-exposure strategies, but it is worth noting that the alternative—attempting to intensively train captive predator-naïve animals to fear predators (as we have done here)—involves skirting a fine line between a training regime that is powerful enough to produce a lasting change in an animal’s behavior and one that is within the bounds of social and ethical licensing. It is possible, for example, that a less abstracted training regime (e.g., placing a muzzled dingo into the quoll enclosures to simulate a genuine predation attempt) may have been more effective in our case (van Heezik et al. 1999; West et al. 2018; Blumstein et al. 2019; Ross et al. 2019). But such a manipulation—conducted on captively maintained animals at an establishment whose primary aim is ethical animal husbandry (e.g., wildlife parks and zoos)—did not win social license and was not approved. In contrast, exposing our havened populations to mild but real predation pressure seems less intensive, less ethically fraught, and, in the long run is likely to be more effective.
Our results show that, even with well-understood teaching techniques, it can be difficult to teach predator recognition to individuals from predator-naïve populations. We tested only a single teaching technique, it therefore is plausible that different techniques may have had greater success in our system. We report our failed training effort here primarily because it is important to document failures so as to prevent duplicate effort, and to avoid biasing the literature. Our results, however, speak more broadly to the problem of predator naïveté in predator-free havens. It would be best if we simply were not in a position of having to re-instill lost antipredator traits, but this situation is an outcome of well-intentioned, and often desperately necessary, conservation efforts. Once the basic capacity to recognize predators is lost in havened or captive zoo populations, it could well be near impossible to train individuals to avoid predators using techniques involving any level of abstraction. Generally, then, it may be both necessary and prudent to expose havened populations to mild predation pressure to prevent loss of antipredator traits. Such an action provides real predation experiences to individuals, but also imposes the selection regime necessary to have populations adapt to conditions outside of predator-free havens. Such an action renders intensive training, such as we attempted here, unnecessary. The conservation community needs to consider the ethics and logistics of such an approach, but it is worth considering that the alternative—intensive training of individuals—may often involve equally difficult ethical considerations with less guarantee of success.
Acknowledgments
Quolls were collected with assistance from Northern Territory Department of Environment and Natural Resources, Kakadu National Park, Territory Wildlife Park, The Endeavour Trust, The Tropical Wildlife Conservancy of North Queensland, and Ella Kelly. Territory Wildlife Park hosted the quolls. Matt Clancy, Alana de Laive, and Ella Kelly, assisted in the execution of this study. Thanks to Bingo the dingo and Lenny the cattledog for attempting to train the quolls, and Tom Parkin and Jas Gray for allowing Lenny and Bingo to participate in the study. Thanks to Alana de Laive for graphic design. We also thank two anonymous reviewers and Chris Pavey for feedback that greatly improved the manuscript. This research was funded by an Australian Research Council Linkage Grant (JKW and BLP LP150100722). CJJ was supported by an Australian Postgraduate Award and David Hay Postgraduate Writing Up Award.
